» 84 - 19 7 0 8

NASA Technical Memorandum 83584

High Efficiency IMP ATT Diodes for 60 GHzIntersatellite Link Applications

Edward J. HauglandLewis Research CenterCleveland, Ohio

Prepared for theThird Communications Satellite Systems Conferencesponsored by the American Institute of Aeronautics and AstronauticsOrlando, Florida, March 18-22, 1984

NASA

https://ntrs.nasa.gov/search.jsp?R=19840011640 2018-05-21T08:41:34+00:00Z

HIGH EFFICIENCY IMPATT DIODES FOR 60 GHz INTERSATELLITE LINK APPLICATIONS

Edward J. Haugland

National Aeronautics and Space AdministrationLewis Research CenterCleveland. Ohio 44135

SUMMARY

IntersatelUte links are expected to play an Increasingly Important role1n future satellite systems. Improved components are required to properlyutilize the wide bandwidth allocated for IntersatelUte link applicationsaround 60 GHz. IMPATT diodes offer the highest potential performance as solid-state power sources for a 60 GHz transmitter. Presently available devices donot have the desired power and efficiency. High efficiency, high power IMPATTdiodes for IntersatelUte link applications are being developed by NASA andother government agencies. This paper describes the development of high effl-.dency 60 GHz IMPATT diodes by NASA. These programs are cofunded by the U.S.Air Force, Space Division.

INTRODUCTION

As satellite systems grow In capacity and complexity, IntersatelUte com-munication links are expected to play an Increasingly significant role. TheNASA Tracking and Data Relay Satellite System (TDRSS) has S-band and Ku-bandIntersatelUte links with low earth orbiting spacecraft as part of Its opera-tional capability. Future NASA, military, and commercial spacecraft willrequire high data rate systems. Nearly 10 GHz of bandwidth Is presentlyallocated around 60 GHz (V-band) for IntersatelUte link applications. Inorder to effectively utilize this resource, electronic components must bedeveloped for Integration Into systems. This paper addresses the developmentof one of those components, namely, an efficient solid-state source of rf powerat 60 GHz.

Of the presently known sol Id-state devices, the IMPATT diode Is the leader1n both theoretical and achieved power from 10 to 300 GHz. The output powervaries as 1/f at lower frequencies due to thermal limitations and then dropsoff as 1/f2 or faster at mm-wave frequencies. This may be seen In figure 1where recent results of Si and GaAs devices are compared (ref. 1). It Is seenthat GaAs has a performance advantage 1n both power and efficiency below 50GHz with Si having a large advantage at 94 GHz and above. The nature of thedegradation observed in GaAs at higher frequencies is uncertain and Is pre-sently under debate. However, very recent work at 60 GHz has decreased theperformance difference and indicates that GaAs 1s likely to have higher powerand efficiency than S1, notwithstanding ongoing work in Si technology. Workat 94 GHz should also be very Interesting with regard to the status of mm-waveperformance of GaAs.

NASA Lewis is currently sponsoring two development contracts for a onewatt, 15 percent efficient, high reliability 60 GHz IMPATT diode as shown Intable 1. The contractors are Hughes Aircraft Company, Electron Dynamics

Division, Torrance, California, and M/A-COM, Gallium Arsenide Products Inc.,Burlington, Massachusetts. Each contractor has chosen GaAs as the activematerial and Is utilizing large signal (nonlinear) computer design techniques.For material growth, Hughes Is using Molecular Beam Epitaxy (MBE) whereas M/A-COM 1s using Organo-MetalUc Chemical Vapor Deposition (MOCVD). These advancedmaterial growth techniques are necessary due to the high frequency and highperformance goals of the program. Each contractor Is utilizing diamond heatsink (DHS) technology due to stringent thermal requirements. After a briefreview of IMPATT operation, the current status of these programs will be dis-cussed In detail.

60 GHz IMPATT OVERVIEW

IMPATT oscillations were first discovered 1n 1965. The field 1s matureand also very active (ref. 2). Present work 1s largely Involved with Improve-ments In performance through refinements In design, material growth, and pro-cessing. The physical picture of the rf power generation 1s straightforward.A pn junction diode 1s reverse biased near breakdown. An rf voltage 1sapplied to the diode. During the half cycle when the rf voltage adds to thedc voltage, electron-hole pairs are created by avalanche Impact lonlzatlon Inthe avalanche region around the junction. Due to the mechanics of this gener-ation process, the maximum charge pulse occurs when the rf voltage has de-creased to zero. This Introduces a 90° phase delay between the rf voltage andthe pulse of charge. The charge pulses then drift toward the diode contactareas. The electrons drift In the n region and the holes 1n the p region,Inducing an external current. During this portion of the rf cycle, the rfvoltage has the opposite sense to the dc voltage. If the diode Is designed sothat the time of transit of the charge pulses 1s equal to one-half of the rfcycle, an additional 90° phase delay 1s Introduced. The situation Is thatthere Is nearly zero current during the positive half of the rf cycle andpositive carrier flow against a negative rf voltage 1n the second half of thecycle, thus generating rf power. For high efficiency and high power, one needsto generate as large a charge pulse as possible while maintaining the properphase relations between current and voltage and also achieving the necessaryImpedance and thermal values.

An example of a 60 GHz GaAs IMPATT diode design will be given. Thisdesign was generated and analyzed by Mains and Haddad (ref. 3) of the Univer-sity of Michigan through the use of their large signal (nonlinear) computersimulation program. The doping profile Is shown 1n figure 2 and the physicaland operating characteristics are listed In table 2. The profile shown Is ahybrid double-drift profile which means that both the p and n materialhave drift regions and that the n material also has a higher doped region atthe pn junction to confine the avalanche. There are several features to benoted. The doping profile requires several sharp changes In a distance ofless than 1 ym. Nonabrupt doping changes and nonunlform doping within layersare known to degrade efficiency. Thus, stringent requirements are placed onthe material growth process. The physical size of the diode Is small and thecurrent density 1s large. This places demands on the entire processing andfabrication sequence. The thermal resistance 1s calculated assuming that adiamond heat sink 1s used, that there 1s a minimum of material between thejunction and the heat sink, and that there 1s no extra thermal resistance frompoor processing. This means that a reliable diamond heat sink technology 1snecessary. Finally, the negative resistance of 1 ohm and capacltlve reactance

of 11 ohms are about the minimum values which may be matched Into rf circuitry.This design was based on the drift-diffusion model of carrier transport 1n asemiconductor. Using a more realistic energy and momentum relaxation model,relaxation effects which occur on the picosecond time scale may be Importantand may change the design features to some degree (ref. 4).

HUGHES APPROACH

In order to achieve the goals listed 1n table 1, Hughes has chosen GaAsusing MBE as their primary material growth approach with conventional VaporPhase Epitaxy (VPE) as a backup. The high frequency and high power goalsrequire several sharp doping changes In a distance less than 1 pm. The highefficiency goal requires that these doping changes be sharp and that dopingwithin each region be very uniform. From an analysis of large signal computersimulations, Hughes has chosen the double-drift hybrid profile. A recentSecondary Ion Mass Spectroscopy (SIMS) analysis of this profile which was grownusing MBE by Perkln-Elmer Corp. 1s shown In figure 3. The sharp transitionsand excellent uniformity within the layers may be seen. The transitions fromone doping level to another take place 1n approximately 200 A distance and thisshould be sufficient for this program. A portion of this apparent transitionwidth 1s due to Instrumental rounding 1n the SIMS analysis. Double-drift flatprofiles have also been grown but have not performed as well. Single-driftprofiles using Schottky contacts were grown early 1n the program but sufferedfrom large leakage and burn out at low bias. MBE material Is also being grownfor Hughes at Cornell University. Installation of a MBE machine at Hughes hasrecently been completed and IMPATT material growth will begin soon. VPE growthhas also been done but has not been as successful. Future plans are to concen-trate exclusively on MBE growth.

Although the Initial diode design was performed using small signal analy-sis, Hughes has worked with UCLA and has set up their own program for largesignal computer analysis. This analysis Is essential for high performance dueto the nonlinear nature of IMPATT operation. There are still questions to beanswered regarding the modelling of parameters which are critical to IMPATToperation. The most Important of these are; the lonlzatlon rate versus elec-tric field, saturation velocity versus temperature, and diffusion constantversus electric field. Recent large signal results show excellent agreementwith experiment In power and efficiency.

Processing of the diodes Is a key step 1n the fabrication sequence. Itwas predicted and experimentally verified that plated heat sink diodes wouldnot provide adequate heat removal from the junction. Hughes has developed apill diode process for thinning and metallizing the diodes for mounting on adiamond heat sink. The pill process has an extra advantage In that the numberof diodes 1s Increased by a factor of 5 over that from plated heat sink pro-cessing. This 1s especially Important due to the small area wafers which areoften necessary 1n MBE machines. During the process, the p metallization(Au/Zn) 1s done followed by wafer thinning from 250 to 10 \>m. The n metal-lization (Au/Ge/N1) 1s evaporated and alloyed, different size dots are definedon p and n sides and the thinned wafer 1s etched through to form the pilldiodes. The metallization 1n this process 1s critical In that the pmetallization must be able to bond well to the metallized diamond withoutexcessive pressure or temperature. This process cured problems associated

with excessive bonding pressures which often lead to cracking of the GaAsdiode. A metallized diamond Is hot pressed Into a copper stud and the pilldiode 1s thermo-compresslon bonded to the diamond. These two steps determinethe thermal characteristics of the device and the quality and reproduclblHtymust be high. A metallized quartz ring and preformed ribbon are used to makeelectrical connection and also to control the magnitude of the parasitic capac-itance and Inductance. Finally, the diode 1s tested 1n a coaxlally-coupledreduced height waveguide cavity. This cavity was chosen for Its flexibility1n changing dimensions and Impedances and also the ease with which devices maybe Inserted and removed without disturbing the circuit.

Although hampered somewhat by a lack of material, good progress Is beingmade. The material problem was due to older MBE machines with very limitedgrowth area per run, -1.5 cm^, and the fact that the difficult doping profilewas not easily achieved. New generation MBE machines which now have excellentuniformity over a 2 1n. wafer and also allow for multiple runs with one loadingwill alleviate the first problem. Operator experience and Increased capabilityof the machines will make the realization of a particular profile easier toaccomplish. Best results to date have been 1.0 W with 13 percent efficiencyat 56 GHz. The noise characteristics will be measured later 1n the program.Thermal resistance values are about 50° C/W and will have to Improve In orderto achieve the desired performance.

M/A-COM APPROACH

For their development of a 60 GHz IMPATT, M/A-COM has chosen GaAs with adouble-drift hybrid Read design and diamond heat sink. The material 1s grownboth by conventional VPE methods and also by the MOCVO technique. The dopingprofile has been obtained through use of the large signal program at theUniversity of Michigan. Several options for 60 GHz operation were analyzedand a double-drift hybrid Read design was chosen. The predicted performancewas 1.3 W output at 14 percent efficiency with a current density of 18kA/cm^ and required thermal resistance of 28° C/W as shown In table 2.

The material growth has proceeded well 1n both VPE and MOCVO. Olethylzinc (DEZ) has been successfully used as a p dopant 1n the VPE growth withgood reprodudblllty. Also, thin layers and sharp transitions have been suc-cessfully demonstrated. The transition sharpness has been observed to be lessthan 500 K using C-V measurements but this Is about the sensitivity of the C-Vtechnique. Nevertheless, due to the rapid growth rate of the VPE technique,1t 1s difficult to execute the many doping changes and to maintain good uni-formity within layers and sharp transitions between layers.

MOCVO, on the other hand, has the potential to grow sharper profiles whilemaintaining good uniformity. MOCVD growth takes place 1n a cold wall reactorby the Irreversible pyrolysls of an alkyl gallium compound and arslne.Trlmethyl gallium (TMG) Is the compound normally used. This pyrolysls takesplace 1n a single hot zone which 1s heated by rf Induction. Low substrategrowth temperature, 650° C, and high gas flow rates permit rapid changes ofdoping level and give this process the capability of producing complex struc-tures with sharp transitions. The entire system has been automated for repro-duclblllty. The dopants being used are sllane and dimethyl zinc (DMZ).

As a result of experiments with baffles and flow rates, uniformity of dopingand thickness 1n the n layer Is ±10 percent run to run and the n layer doping1s uniform to ±10 percent across a 2" wafer. There 1s more uncertainty 1n thep layer measurements and these values may be ±20 percent. Work 1s continuing1n that area. It Is felt that the doping transitions are sharper than 200 Aalthough this 1s below the sensitivity of the C-V measurements.

The rf testing Is accomplished using a "Top-Hat" circuit. This circuitwas chosen for Its broadband capability and for ease of Interchanging diodes.Impedance matching 1s accomplished by a variety of top-hats with differentgeometries which are easily Interchanged and also by Interchangeable heatsinks of different sizes. This circuit has proven to be versatile In testingdiodes with widely varying characteristics. Best rf results to date have beenwith VPE grown double-drift diodes. Highest power has been 300 mW with 11percent efficiency at 52 GHz. HOCVD results have been less than 10 mW withlow efficiency. The cause of this low output has not been determined andadditional experiments are being done to determine the causes. Plated heatsinks are being used for this Initial evaluation. The need for diamond heatsink to attain the 1 W level Is recognized and work Is 1n progress to developthis technology for millimeter wave devices.

RELATED 60 GHz IMPAT1 DIODE WORK

Due to Interest 1n IntersatelUte link applications and secure terrestriallinks, there are several 60 GHz IMPATT diode programs 1n progress. They arelisted 1n table 3. The Hughes-NASA/GSFC program 1s specifically designed toImprove the mature S1 technology through processing, fabrication, and designoptimization. The Var1an-NRl program with InP has experienced problems withthe p-layer and the results have been less than expected. Additional workwill continue 1n order to explore the potential of InP for high efficiency.The remaining four programs with LeRC and AFWAL all have similar goals withGaAs but utilize different approaches. These programs ought to clearly Indi-cate whether the knee In the curves of figure 1 Is above or below 60 GHz. Thetrend of the results from Hughes and from Raytheon would Indicate that the1/f2 falloff 1s going to take place above 60 GHz. As part of the same pro-gram, Raytheon 1s also doing CW GaAs IMPATT work at 94 GHz and that data pointfor the GaAs-S1 comparison will be significant.

The question of reliability Is, of course, extremely Important. Masseand Harper recently summarized the results of published reliability data onGaAs IMPATT devices (ref. 5). At a given junction temperature, say 250° C,the median times to failure varied over 5 orders of magnitude, depending onthe metallization used and types of stress employed. The reliability ofIMPATT devices depends basically on the operational junction temperature andthe metallization system used. Reliability testing will be done at theconclusion of the NASA programs and separate reliability programs may benecessary. Most estimates Indicate, however, that a 10 year lifetime In spaceshould be achievable with junction temperatures below 250° C.

CONCLUDING REMARKS

For the NASA Tracking and Data Acquisition System (TDAS) of the early1990's time frame, system studies Indicate that V-band 1ntersatell1te linkswill require approximately 10 W of transmitter power. Table 4 shows recentand future 60 GHz amplifier developments. The new development will probablyrequire 75-30 devices depending on the mode of operation chosen and thedevices available. Presently available devices are 1n the 0.8 M range with6-9 percent efficiency. With the higher power, higher efficiency devices underdevelopment, there will be substantial savings 1n amplifier complexity, weight,and prime power requirements. Weight and power may be reduced by 30-40 percentof values 1n table 4. This Is expected to have a strong Impact on the viabil-ity of the solid-state amplifier for IntersatelUte links.

Optical and travelling wave tube (TWT) technology offer alternatives tothe solid-state amplifier described above. Work Is underway 1n both areas.At the present time, the optical system weight 1s substantially higher thanthat required for a 60 GHz solid-state system. TWT's offer substantially morepower and efficiency than solid-state amplifiers. A 75 W TWT with 40 percentefficiency Is presently being developed at 60 GHz. However, at the power levelof 10 W, an IHPATT based sol Id-state amplifier may be competitive with theTWT. Efficient devices and power combining schemes may bring the solid-stateamplifier weights down to comparable TWT values. Power supplies forsolid-state amplifiers are low voltage and less complex. Reliability of theIMPATTS 1s a key question which must be addressed.

In summary, the application of sophisticated computer design techniques,advanced material growth techniques, and careful thermal considerations areexpected to lead to high efficiency high power IMPATT sources of power at 60GHz. These devices are expected to find use for IntersatelUte link trans-mitters in the 1990's.

REFERENCES

1. Hleslmalr H., DeSantls C., and Wilson N., "State of the Art of Solld-Stateand Tube Transmitters," Microwave Journal. Vol. 26 No. 10, Oct. 1983,pp. 46, 48.

2. "IMPATT and Related Transit-Time Diodes," Physics of SemiconductorDevices^ S. M. Sze, 2nd ed., John Wiley and Sons, 1981, pp. 566-636.

3. Mains R. K., and Haddad G., "Capabilities and Potential of Millimeter-WaveIMPATT Devices," AFWAL-TR-82-1141, Nov. 1982.

4. Mains R., Haddad, G., and Blakey P., "Simulation of GaAs IMPATT DiodesIncluding Energy and Velocity Transport Equations," IEEE Transactions onElectron Devices. Vol. ED-30, No. 10 Oct. 1983, pp. 1327-1338.

5. Masse D., and Harper R., "Solid State Amplifiers for Satellite Communi-cations," AIAA Paper 82-0547, 1982.

TABLE I. - 60 GHz IMPATT DIODENASA/LEWIS RESEARCH CENTER

Goals

Operable frequency, GHz

Output power (Oscillator), watt CW . .

Reliability

Contractors

Hughes Aircraft Co., Electron DynamicsTorrance, CA

M/A-COM Gallium Arsenide Products, IncBurlington, MA

59 61

1

. 15

Compatible with 10year lifetime inspace

Div.,

TABLE II. - 60 GHz GaAs HYBRID DOUBLE DRIFTIMPATT DIODE CHARACTERISTICS FROM

LARGE SIGNAL COMPUTER DESIGN (REF. 3)

Diameter -

Area

JDC

JDC

VDC

PDC

2 mil

2.03xlO-5cm2

17.7 kA/cm2

359 mA

25.5 V

9.3 W

Prf -

n

A T j -

6

GD -

BD -

1.3 W

14.2 percent

225° C

a28.2° C/W

(-)8.04 mS

89.3 mS

aWith diamond heat sink and metallizationcontribution of 5.0° C/W.

TABLE III. - 60 GHz IMPATT DIODE PROGRAMS

Contractor

Hughes

M/A-COM

Raytheon3

Hughes

Varian

TRua,b

Sponsor

NASA/LeRC

NASA/LeRC

AFWAL

NASA/GSFC

NRL

npUAi

Goals

1 W 15 percent

1 W 15 percent

2 W 15 percent

1.5 W 14 percent

1 W 14 percent

Approach

GaAs/MBE/DHS

GaAs/MOCVD/DHS

GaAs/MBE/DHS

Si/VPE/DHS

InP/VPE

RaA<;/MRF

Results (12/83)

1.03 W at 13 percent

0.3 W at 11 percent

1.25 W at 11 percent

1.6 W at 10 percent

1.7 W at 4 percent(1 percent Dutycycle)

^Multiple frequency programs."Program requires substantial university participation.

TABLE IV. - 60 GHz AMPLIFIERS SPONSOREDBY NASA/GSFC

Contractor

CW power out(W)

Gain (dB)

Bandwidth (GHz)

Efficiency (percent)

Prime power req. (W)

Device

Estimated flight Wt (kg)(with regulators)

Completion date

Phase I

Hughes,

goals/results

4/3.6

30/30

2.5/2.5

4/3

100/120

10 Si IMPATTS,7 percentefficient

6

1983

Phase II

to be determined,

goals

10

40

1

6

170

-_— —

13

1985

100 rr- 100 p—

10

oLU0.

O

.1

.01

10

oa.

.01

Si

GaAs

O GaAs• Si

POWEREFFICIENCY

I I I . I.I I . I. I. I.I I1 2 4 6 10 20 4060100 200 400

FREQUENCY, GHz

Figure 1. - Power and efficiency data for GaAs and Si IMPATTdiodes (after ref. 1).

10lvi—

1018

0

Q-o

1017 -

1016l_

dc

^^^

_

P-«—

PING PROFILEELECTRIC FIELD

n— »-

A//

/1I

/*

/t

(

— i

\\\\\•\\

0 .2 .4

-^^^^^

^^

1 1

~̂™"

-

1

8

7

6e0

5 ]>

ca4 d

u.O

3 p£0aUJ

2

1

0.6 .8 1.0

DISTANCE, urn

Figure 2. - 60 GHz GaAs hybrid double-driftdoping profile and dc electric field at JDC •18kA/cm2. (After ref. 3.)

Oz8o

10,19

1018

1017

,1610

1015

Si

1 -DESIGN PROFILE |"

Be

I I 1 1 1 1

I.5 1.0

DEPTH, Mm1.5 2.0

Figure 3. - Doping profile of 60 GHz IMPATT material grownby MBE and measured by SIMS compared to design profile.The p dopant is Be and the n dopant is Si.

1. Report No.

NASA TM-83584

2. Government Accession No 3. Recipient's Catalog No.

4. Title and Subtitle

High Efficiency IMPATT Diodes for 60 GHz Intersatellite

Link Applications

5. Report Date

6 Performing Organization Code

506-58-227. Author(s) 8. Performing Organization Report No

E-1974Edward J. Haugland 10 Work Unit No

9 Performing Organization Name and Address

National Aeronautics and Space AdministrationLewis Research CenterCleveland, Ohio 44135

11 Contract or Grant No

12. Sponsoring Agency Name and Address

National Aeronautics and Space Administration

Washington, D.C. 20546

13. Type of Report and Period Covered

Technical Memorandum14. Sponsoring Agency Code

15 Supplementary Notes

Prepared for the Tenth Communications Satellite Systems Conference sponsored by theAmerican Institute of Aeronautics and Astronautics, Orlando, Florida, March 18-22,1984. This research was cofunded by the U.S. Air Force, Space Division.

16. Abstract

Intersatellite links are expected to play an increasingly important role infuture satellite systems. Improved components are required to properly utilizethe wide bandwidth allocated for intersatellite link applications around 60 GHz.IMPATT diodes offer the highest potential performance as solid state powersources for a 60 GHz transmitter. Presently available devices do not have thedesired power and efficiency. High efficiency, high power IMPATT diodes forintersatellite link applications are being developed by NASA and other governmentagencies. This paper describes the development of high efficiency 60 GHz IMPATTdiodes by NASA.

17. Key Words (Suggested by Authors))

60 GHz; GaAs; IMPATT; Molecular beamepitaxy; Organo-metallic chemical vapordeposition; Diamond heat sink

18. Distribution Statement

Unclassified - unlimitedSTAR Category 33

19 Security Classif. (of this report)

Unclassified

20. Security Classif. (of this page)

Unclassified

21 No. of pages 22 Price'

*For sale by the National Technical Information Service, Springfield, Virginia 22161

National Aeronautics andSpace Administration

Washington, D.C.20546

Official Business

Penalty for Private Use. $300

SPECIAL FOURTH CLASS MAILBOOK

Postage and Fees PaidNational Aeronautics andSpace AdministrationNASA-451

NASA POSTMASTER: If Undeliverahle (Section I SKPostal Manual) Do Not Return